HGF and c-Met Participate in Paracrine Tumorigenic Pathways in Head and Neck Squamous Cell Cancer

Purpose: We determined hepatocyte growth factor (HGF) and c-Met expression and signaling in human head and neck squamous cell carcinoma (HNSCC) cells and primary tissues and tested the ability of c-Met tyrosine kinase inhibitors (TKI) to block HGF-induced biological signaling.

Experimental Design: Expression and signaling were determined using immunoblotting, ELISA, and immunohistochemistry. Biological end points included wound healing, cell proliferation, and invasion. c-Met TKIs were tested for their ability to block HGF-induced signaling and biological effects in vitro and in xenografts established in nude mice.

Results: c-Met was expressed and functional in HNSCC cells. HGF was secreted by HNSCC tumor-derived fibroblasts, but not by HNSCC cells. Activation of c-Met promoted phosphorylation of AKT and mitogen-activated protein kinase as well as release of the inflammatory cytokine interleukin-8. Cell growth and wound healing were also stimulated by HGF. c-Met TKIs blocked HGF-induced signaling, interleukin-8 release, and wound healing. Enhanced invasion of HNSCC cells induced by the presence of tumor-derived fibroblasts was completely blocked with a HGF-neutralizing antibody. PF-2341066, a c-Met TKI, caused a 50% inhibition of HNSCC tumor growth in vivo with decreased proliferation and increased apoptosis within the tumors. In HNSCC tumor tissues, both HGF and c-Met protein were increased compared with expression in normal mucosa.

Conclusions: These results show that HGF acts mainly as a paracrine factor in HNSCC cells, the HGF/c-Met pathway is frequently up-regulated and functional in HNSCC, and a clinically relevant c-Met TKI shows antitumor activity in vivo. Blocking the HGF/c-Met pathway may be clinically useful for the treatment of HNSCC.

c-Met, the tyrosine kinase receptor for hepatocyte growth factor (HGF), is overexpressed in a variety of tumors in which it plays a central role in malignant transformation (1). Activation of c-Met by HGF leads to a variety of cellular signals that mediate tumor growth, metastasis, and angiogenesis (1, 2). Important downstream signals of c-Met include p44/p42 mitogen-activated protein kinase (MAPK) and PI3K/AKT, which are critical effectors of tumor cell proliferation and survival, respectively (3). In addition, c-Met has been shown to activate signal transducers and activators of transcription 3 (STAT3) and induce expression of matrix metalloproteinases as well as growth factors (4, 5). Given its critical role in tumor function, c-Met is emerging as a therapeutic target for cancer therapy. Treatment strategies in clinical testing include small molecule inhibitors specific for the tyrosine kinase domain of c-Met and monoclonal antibodies against HGF (6).

Head and neck squamous cell carcinoma (HNSCC) constitutes 90% of all head and neck cancers and is often fatal despite current treatment protocols implementing surgery, radiation, and chemotherapy. Even with the advent of new treatment protocols, no significant improvement in 5-year overall survival rates have been made (7). Novel targeted therapeutic strategies aimed at growth pathways utilized by the tumor cells should have potential to treat this disease. c-Met overexpression has been reported in HNSCC (8, 9). This tumor type has also been shown to amplify epidermal growth factor receptor (EGFR), a receptor tyrosine kinase that shares many important signal intermediates common to the c-Met/HGF pathway such as p44/p42 MAPK, PI3K/AKT, and STAT (10). This signaling redundancy could be the reason for the modest benefits achieved after therapy with the EGFR tyrosine kinase inhibitors (TKI) gefitinib or erlotinib in patients with HNSCC (10). Given the limited therapeutic efficacy found with EGFR-targeted therapy in HNSCC, inhibitors that target other tyrosine kinase receptors exhibiting similar signaling activation, either alone or in combination, may be beneficial.

Here we determined the extent of c-Met expression and signaling through HGF in human HNSCC cells, as well as HGF and c-Met expression in HNSCC primary tumors and normal mucosa. We show that ligand activation of c-Met in HNSCC cells is mainly through a paracrine mechanism whereby the stromal fibroblasts derived from the tumors produce HGF whereas the tumor cells themselves do not. We also used preclinical models to show the antitumor effects of small molecule c-Met TKIs. This study shows that c-Met activation promotes HNSCC cell signaling and protumorigenic cell responses, and that c-Met blockade with a clinically relevant c-Met TKI leads to HNSCC tumor growth suppression in vivo. We also show that HGF and c-Met protein expression are significantly overexpressed in patient tumor samples compared with adjacent noncancerous mucosa. In addition, we show that patients whose tumors have low levels of both proteins have a trend towards better overall survival. Together these data suggest that c-Met pathway targeting may have therapeutic benefits in HNSCC and is a potential new strategy for HNSCC patients resistant to other approaches.

Cell lines. PCI-37A, UM-22B, and 1483 HNSCC cell lines were from human origin as previously described (11–13). Tumor cells were maintained in DMEM (Mediatech, Inc.) supplemented with 10% fetal bovine serum and verified to be mycoplasma-free. Human papilloma virus status was determined by the method of Martinez et al. (14) and found to be negative in all three cell lines (Supplementary Fig. S1).

Stromal fibroblasts were isolated from HNSCC tumors and are referred to as tumor-derived fibroblasts (TDF). Briefly, HNSCC tumor tissue was rinsed in 100× penicillin/streptomycin/ampohotericin B antibiotic cocktail. The tumors were minced and subjected to collagenase II, DNase I, and trypsin treatment at 37°C for 15 min. The cells were washed and plated in DMEM containing 10% fetal bovine serum and incubated at 37°C in a 5% CO₂ incubator. When confluent, fibroblasts were trypsinized and used in the experiments. Primary fibroblasts were propagated in culture for up to 20 passages. All cells were confirmed to be mycoplasma-free. Three different fibroblast cultures were utilized in this study: TDF 0001, TDF 0731, and TDF 0733, and originate from tumors of the tongue, larynx, and tonsil, respectively.

Pharmaceutical reagents. SU11274 was purchased from EMD Chemicals, Inc. PF-2341066 was generously provided by Pfizer. Recombinant human HGF and HGF-neutralizing antibody (HGF NA) were purchased from R&D Systems and prepared according to the manufacturer's instructions. SU11274 and PF-2341066 were dissolved in DMSO and stored at −20°C for use in vitro. PF-2341066 was dissolved in water for injection into animals.

Immunoblot analysis. Cells were lysed in 2× SDS sample buffer or ice-cold radioimmunoprecipitation buffer [10 mmol/L Tris-HCl (pH 7.6), 50 mmol/L NaCl, 5 mmol/L EDTA, 50 mmol/L NaF, 30 mmol/L sodium pyrophosphate, 1 mmol/L sodium vanadate and protease minitab; Roche Applied Science], harvested using a cell scraper and passed 10 times through an 18-gauge needle. Insoluble material was removed by centrifugation (14,000 rpm) for 10 min at 4°C. Protein was separated by SDS-PAGE, transferred onto nitrocellulose, and probed overnight at 4°C with anti-Met (Santa Cruz Biotechnology), anti-p-Met (Tyr1234/1235), anti-MAPK, anti-p-MAPK (Th202/Tyr204), anti-AKT, anti-p-AKT (Ser473; Cell Signaling Technology), and anti-actin (Millipore). Immunoreactivity was detected using antimouse or antirabbit IgG conjugated peroxidase and visualized by enhanced chemiluminescence. Quantitation of immunoreactive signals was done by densitometry and Molecular Dynamics ImageQuant software analysis (Version 5.2). The ratio of activated protein form to total protein form was calculated and expressed relative to their respective controls for each experiment.

Scratch wound migration assay. Confluent HNSCC cells were serum-deprived for 48 h and wounds were generated using a sterile 200-μL pipet tip. Cells were then exposed to specified treatments and grown for an additional 48 h. Wound closure was assessed using an Olympus IX71 at 10× magnification. Cell migration distance was measured using Adobe Photoshop 9.0.2 software (Adobe Systems Incorporated) and compared with baseline measurements.

Cell proliferation assay. HNSCC cells were plated in 96-well plates at 6 × 10³ cells/well and grown in Basal Medium Eagle's (Invitrogen Life Technologies, Inc.) supplemented with 1% fetal bovine serum (Atlanta Biologicals) and 2 mmol/L L-glutamine (Invitrogen) for 24 h prior to experimental treatments. Cell proliferation was monitored after 72 h using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay Kit (Promega Corporation). Metabolically active cells were labeled with MTS tetrazolium for 1.5 h and measured at 490 nm using an absorbance plate reader (Bio-Rad Laboratories).

Invasion assay. Cell invasiveness was evaluated in vitro using Matrigel-coated modified Boyden chamber inserts with a pore size of 8 μm (Becton Dickenson/Biocoat). HNSCC cells were plated at a density of 5 × 10³ cells in the insert. TDF 0001 cells were plated in the lower well (2 × 10⁴ cells/well). Both inserts and lower wells were treated with either the vehicle control (DMEM), HGF NA (30 ng/mL), or control antibody. After 24 h of treatment at 37°C in a 5% CO₂ incubator, the cells in the insert were gently removed by using a cotton swab. Cells on the insert's reverse side were fixed and stained with Hema 3 (Fisher Scientific) according to the manufacturer's instructions. In the four representative fields, invading cells were counted using light microscopy at 400× magnification. Mean ± SE was calculated from two independent experiments.

HNSCC xenografts and sensitivity to c-Met inhibition. UM-22B tumor cells (3 × 10⁶) were injected s.c into the flanks of nude mice. The mice were randomized into two treatment groups with eight animals per group. PF-2341066 was administered at 12.5 mg/kg/d by oral gavage. Treatment started 7 d following tumor inoculation. Tumor size was measured two times per week and reported as tumor volume (mm³). Animal care was in strict compliance with the institutional guidelines established by the University of Pittsburgh.

At the end of the treatment period, the animals were sacrificed and the tumors were removed and fixed in 10% buffered formalin for immunohistochemical analysis. Formalin-fixed tumors were paraffin-embedded,sliced, and mounted on slides. Paraffin was removed from the slides with xylenes and slides were stained with H&E to examine the tumor morphology. For the apoptosis assay, the number of apoptotic cells was determined using the ApopTag Peroxidase In Situ Apoptosis Detection Kit (Millipore) as described previously (15). Brown staining was considered positive. Slides were read and scored for the number of positive tumor cells per five high powered fields per sample. Results are reported as the mean ± SE.

Human tissue samples and immunohistochemistry. Tissues were collected under the auspices of a tissue bank protocol approved by the University of Pittsburgh Institutional Review Board. Two tissue microarrays were constructed using tumor specimens from 56 HNSCC patients who underwent surgical resection with curative intent, 26 with paired adjacent histologically confirmed normal mucosa. Triplicate 6-mm cores were extracted from paraffin-embedded tissue blocks from each surgical specimen and arrayed on two recipient paraffin blocks. The newly constructed arrays were then warmed to 37°C for 10 min to allow annealing of donor cores to the paraffin wax of the recipient block. For tissue microarray quality assessment and morphologic confirmation of tumor, one H&E-stained slide was evaluated for every ten tissue sections. Presence of tumor or histologically normal mucosa within the tissue cores was confirmed by a head and neck cancer pathologist (RS).

Tissue microarray sections were deparaffinized with xylene and ethanol. Tissue HGF and c-Met protein expression were measured by staining tissue microarrays with anti-HGF (R&D Systems; 1:200) or anti-c-Met (Santa Cruz Biotechnology; 1:75). HGF immunoreactivity was detected using biotinylated goat/rabbit IgG secondary antibody (Dako) and Envision Rabbit Polymer (Dako). c-Met signal amplification was done using antibody-conjugated proprietary micropolymer peroxidase (ImmPRESS, Vector labs). Immunoreactive cells were visualized following incubation with diaminobenzidine chromogenic substrate at room temperature for 4 to 5 min. Sections were counterstained with hematoxylin for 2 to 2.5 min to provide morphologic detail. Immunohistochemical staining of each core was scored semiquantitatively by a head and neck cancer pathologist. The percentage of immunoreactive cells was recorded and rounded to the nearest 5th percentile. Intensity was scored as 0 (none), 1+ (weak), 2+ (moderate), or 3+ (strong). A composite score, the immunohistochemistry score, was derived from the product of the percentage and intensity of staining, and these composite scores were averaged for the triplicate cores. Images from select patient cores were recorded using 100× magnification. Positive (breast carcinoma tissue) and negative (omission of primary antibody) immunohistochemistry controls were done for both antibodies with expected specificity observed. In addition, these antibodies give specific immunoreactive bands of the correct size in Western blots (16, 17).

ELISA measurements. HNSCC tumor cells and tumor-derived fibroblasts were grown in 10-cm dishes to 70% confluency using DMEM growth medium. Four milliliters of cell culture medium were collected and examined for HGF levels using the Quantikine Human HGF ELISA kit (R&D Systems) according to the manufacturer's instructions.

Human IL-8 Quantikine ELISA kits were purchased from R&D Systems and the manufacturer's instructions were followed. HNSCC cells were grown as described above followed by serum deprivation for 48 h and subsequent treatment with 50 ng/mL HGF for 0 to 48 h or with pretreatment with 2.5 μmol/L SU11274. Cell culture supernatant was removed and analyzed directly for interleukin 8 (IL-8) secretion. Results represent the mean ± SE of four samples per experimental treatment.

Statistical analysis. Biochemical and animal data were analyzed using unpaired two-tailed Mann-Whitney t test or one-way ANOVA followed by Tukey-Kramer multiple comparisons test (Graphpad Instat 3.06, GraphPad Software, Inc.). Comparison of tumor and adjacent muscosa immunohistochemistry scores was carried out using two-tailed Wilcoxon signed-rank test (SPSS V14.0, SPSS Inc.). Correlation between tumor c-Met and HGF protein levels was evaluated using Spearman's correlation coefficient. Overall survival differences between patients with high versus low tumor HGF or c-Met protein levels, as defined by the median immunohistochemistry score for each antigen, were evaluated using Kaplan-Meier plots and the log-rank test for equality of survivors. All statistical tests were two-sided with the threshold for statistical significance defined as P < 0.05.

c-Met is expressed and functional in HNSCC tumor cells. Although c-Met expression has been previously reported in HNSCC, the biological significance of this proto-oncogene in this cancer is largely unknown. To determine if c-Met is expressed in HNSCC cells, we examined expression of total c-Met protein in a panel of HNSCC cell lines in complete media. We found that c-Met protein expression was detected in UM-22B, PCI-37A, and 1483 HNSCC tumor cells (Fig. 1A). To determine the cellular source of the ligand for c-Met, we also examined cell culture media collected from HNSCC cells for released HGF, the c-Met ligand. These HNSCC cell lines did not produce detectable HGF to activate c-Met in an autocrine manner (Fig. 1B). Instead, we observed release of HGF (range 625 ± 35 to 8,115 ± 544 pg/mL) into culture media by three different cultures of stromal fibroblasts derived from the microenvironment of HNSCC tumors, also called TDF cells, suggesting the ligand activation of c-Met is mainly through a paracrine mechanism in HNSCC. It remains a possibility that some autocrine signaling may also occur.

The addition of a minimum of 1 ng/mL recombinant human HGF to cultured HNSCC cells induced c-Met phosphorylation and activated two key downstream effector molecules of HGF/c-Met, AKT and p44/p42 MAPK, in a concentration-dependent fashion (data not shown). The maximal effect (at 50 ng/mL HGF treatment) is shown in three HNSCC cell lines in Fig. 1C. Phospho-c-Met was induced by 5 minutes between 2.7- and 4.3-fold over zero time. Phospho-AKT was also induced within 5 minutes 6- to 7-fold. Due to high basal levels, p44/p42 MAPK was induced by HGF to a lesser extent, 1.3- to 1.7-fold, in all three cell lines, indicating that this end point is a poor readout of HGF activity in these cell lines.

It has been shown previously that activation of the HGF signaling pathway leads to release of cytokines that can promote inflammation, movement of tumor cells through endothelium, and angiogenesis (18). One such cytokine is IL-8. To determine if HGF can induce IL-8 release in HNSCC cells, cells were serum-deprived followed by treatment with HGF over time. Within 15 minutes of HGF treatment (50 ng/mL), and lasting for 48 hours, the IL-8 protein was released, reaching levels of over 2 ng/mL in the culture medium (Fig. 1D), as determined by ELISA. The detected levels of IL-8 were significantly different from control at P < 0.0001 beginning at 1 to 2 hours and continuing to 48 hours. Similar results were observed for release of the angiogenic growth factor, vascular endothelial growth factor (data not shown).

HGF induces wound closure, cell proliferation, and invasion in HNSCC cells. In order to determine the biological relevance of the HGF pathway in HNSCC cells, we examined the effects of exogenous HGF treatment on the ability of HNSCC cells to induce wound closure, proliferation, and cell invasion. Treatment with HGF caused a concentration-dependent increase in tumor cell migration and proliferation, indicating that the c-Met/HGF pathway is functionally relevant in this panel of HNSCC cell lines (Fig. 2A and B). HNSCC cells migrated into an artificially produced wound in the culture dish to a significantly greater extent in the presence of 50 ng/mL HGF in all three HNSCC cell lines. Percent wound closure at 48 hours was 82% to 98% in the presence of HGF, compared with 38% to 50% in vehicle-treated controls (Fig. 2A). A concentration-dependent enhancement of cell proliferation in an MTS assay was also observed with HGF treatment. Beginning at 1 ng/mL, there was an increase in cell proliferation in PCI-37A cells that was maximal at 50 ng/mL HGF (40% over control; the results were statistically significant at 10 and 50 ng/mL). A similar result was seen in UM-22B cells (data not shown).

Because we have shown that stromal cells from the HNSCC microenvironment secrete HGF (Fig. 1B) and it has been reported recently that stromal tissue plays an important role in regulating invasion in oral squamous cell carcinoma (19), we tested the ability of TDFs to increase invasion of HNSCC cells in a three-dimensional coculture assay. TDF cells plated on the bottom of a culture plate induced a statistically significant (P < 0.001) 3.8-fold increase in invasion of HNSCC UM-22B cells plated over the stromal cell layer. A representative experiment is shown in Fig. 2C. These changes could possibly represent a combination of invasion and proliferative effects because a limitation of these types of multiple cell type coculture experiments is that we cannot fully differentiate between these two processes. We next evaluated if released HGF was responsible for the induction of HNSCC cell invasion elicited by stromal cells. Pretreatment with an HGF NA, which was shown previously to block c-Met signaling by >90% (15), completely blocked the ability of TDF cells to enhance invasion of UM-22B cells (Fig. 2C; P < 0.001 TDF versus TDF + anti-HGF antibody). HGF NA had no effect in the absence of TDFs (data not shown) or with pretreatment using a control antibody (Fig. 2C).

c-Met tyrosine kinase inhibitors abrogate the biological responses induced by HGF in HNSCC cells. We next examined the ability of the c-Met small molecule tyrosine kinase inhibitor SU11274 to block c-Met activation in response to HGF. SU11274 has been shown to block c-Met signaling in ovarian cancer, small cell and non–small cell lung cancer, melanoma, and mesothelioma cell lines (20–24), and has an average IC₅₀ on enzymatic kinase activity of purified c-Met of 0.01 μmol/L in vitro (25).

SU11274 pretreatment inhibited c-Met activation in response to 50 ng/mL HGF in a concentration-dependent manner, with complete inhibition of c-Met phosphorylation occurring between 0.5 and 1 μmol/L in three HNSCC cell lines. This inactivation of c-Met by SU11274 coincided with the loss of AKT phosphorylation and a more modest inhibition of p44/p42 MAPK phosphorylation in response to HGF (Fig. 3A). The modest effects observed on MAPK inhibition are a result of the high basal levels of MAPK activation in these cell lines because they produce high levels of EGFR ligands resulting in autocrine activation of EGFR and an alternate route for MAPK activation. Similar concentrations of SU11274 were sufficient to inhibit HNSCC cell migration (Fig. 3B). In a cell migration assay, 1 and 2.5 μmol/L SU11274 significantly reduced the extent of wound closure in UM-22B and 1483 cells in response to HGF and at 2.5 μmol/L in PCI-37A cells. SU11274 also inhibited the growth of all three HNSCC cell lines tested with IC₅₀ values ranging from 4.1 to 5.5 μmol/L in a proliferation assay, and blocked HNSCC invasion into an artificial extracellular matrix (Supplementary Table S1 and data not shown). SU11274 also blocked the HGF induction of IL-8 release by 86% after 1 hour of HGF treatment (Fig. 3C).

SU11274 is not a feasible drug for clinical use because of solubility problems and ulceration effects when used in vivo (26, 27). A newer, more potent, orally bioavailable c-Met TKI, PF-2341066, can be successfully delivered in vivo and has IC₅₀ values ranging from 5 to 20 nmol/L for various biological end points such as c-Met phosphorylation, proliferation, migration, or invasion (26). PF-2341066 showed similar IC₅₀ values (range, 4.0-4.7 μmol/L) to those observed for SU11274 in proliferation assays (Supplementary Table S1). We examined the ability of PF-2341066 to inhibit c-Met signaling and biological effects in UM-22B cells (Fig. 4). PF-2341066 inhibited phosphorylation of c-Met and activation of AKT in response to 50 ng/mL HGF beginning at 1 nmol/L, with complete inhibition at 25 nmol/L (Fig. 4A). No significant inhibition of p44/p42 MAPK was observed, likely due to the high basal levels that were observed in serum-deprived cells (Fig. 1C). Wound closure in UM-22B cells was also significantly inhibited at 250 and 500 nmol/L (Fig. 4B).

c-Met tyrosine kinase inhibitor PF-2341066 inhibits tumor growth in an in vivo tumor xenograft model of HNSCC. To analyze the effects of c-Met inhibition on tumor growth, we tested the effect of PF-2341066 in a murine xenograft model using UM-22B cells that we showed to be responsive to PF-2341066 in vitro (Fig. 4). Daily oral administration of 12.5 mg/kg PF-2341066, a dose used in previous studies by the manufacturer to produce 60% to 70% inhibition of tumor volume in GTL-16 gastric carcinoma cells (which express constitutively active c-Met as a result of c-Met amplification) and U87MG glioblastoma cells (which express both HGF and c-Met; ref. 26) resulted in a 50% reduction in tumor xenograft volume of UM-22B cells after 12 days of treatment (Fig. 5A). H&E staining of tumors at 18 days revealed a highly disorganized structure in PF-2341066–treated tumors compared with vehicle control (Fig. 5B), with distinct evidence of cell debris. Tumors in the PF-2341066 treatment group also contained a significantly higher number of apoptotic cells compared with vehicle control (Fig. 5C).

HGF and c-Met proteins are significantly overexpressed in HNSCC tumors compared with adjacent noncancerous mucosa. To extend these in vitro findings to HNSCC clinical specimens, HGF and c-Met protein levels were evaluated by immunohistochemical analysis in 56 HNSCC patient tumors and 26 paired adjacent histologically normal mucosa in order to determine whether (a) HGF and/or c-Met protein expression differed between patient tumor and normal tissues, and (b) whether HGF and/or c-Met tumor levels had potential utility as prognostic indicators. A summary of HNSCC patient characteristics is provided in Supplementary Table S2. Examples of tissue microarray immunohistochemical staining for HGF and c-Met in tumor and histologically normal adjacent tissues are shown in Fig. 6A. We found that both HGF and c-Met expression were significantly up-regulated in HNSCC tumors compared with levels in the normal mucosa from the margins of the same tumor specimen (HGF, P < 0.001; c-Met, P = 0.04; Fig. 6B). Neither HGF nor c-Met levels were significantly correlated with any subject characteristic or clinical parameter such as sex, age, tumor pathology, or tumor site. HGF and c-Met were detected in the majority of HNSCC tumors evaluated, 82% and 77%, respectively. Frequency distributions of immunohistochemistry scores for tumor HGF and c-Met immunohistochemistry scores indicate that HGF was more frequently present in HNSCC tumors at higher levels than c-Met (Fig. 6C). Tumor levels of c-Met and HGF were not correlated (Spearman's ρ = 0.06; P = 0.65). Plasma was available from 21 of the HNSCC patients used in our immunohistochemical analysis and there were also no correlations observed between circulating HGF levels found in patient plasma and tumor immunohistochemical levels from the same patient (data not shown).

We next evaluated differences in overall survival between patients with high versus low tumor levels of HGF and/or c-Met in order to assess whether HGF and/or c-Met tumor levels were associated with prognosis in our cohort. At last follow-up, 28 patients had died. Median survival for those patients who had died was 14.2 months (range, 1.5-81.0 months); median survival for those patients alive at last follow-up was 42.9 months (range, 17.6-82.6 months). Survival did not differ between subjects with low versus high tumor levels of HGF or c-Met, P = 0.19 and P = 0.58, respectively. Kaplan-Meier plots for survival by HGF and c-Met tumor levels are provided in Supplementary Data Fig. S2A and B, respectively. However, patients with tumors that expressed low levels of both HGF and c-Met tended to have better overall survival than patients with tumors that expressed high levels of HGF, c-Met, or both (Supplementary Fig. S2C). However, this difference was not statistically significant at 95% confidence (P = 0.09; log rank test).

Activation of the HGF/c-Met signaling axis produces significant biological effects in human HNSCC. These effects include stimulation of signaling molecules important in cell proliferation and survival, release of cytokines that induce inflammation and angiogenesis, cell migration, and invasion. Overexpression of either c-Met or HGF, or both, was found in HNSCC tumor tissues, and evidence for a paracrine mechanism of action was found in cells isolated from HNSCC tumors, with production of HGF observed in TDFs derived from HNSCC tumor tissue and c-Met signaling activated in HNSCC epithelial tumor cells. These findings suggest that the signaling cascade induced by the c-Met pathway plays an important role in the growth and survival of human HNSCC, and contributes to invasion into normal tissues, neovascularization, and metastasis. All of these attributes suggest the HGF/c-Met pathway may replace or augment signaling pathways induced by other tyrosine kinase receptors, such as the EGFR, making it a good target for therapy in HNSCC. Inhibitor studies showed that c-Met TKIs could inhibit HGF-induced phosphorylation, signal transduction, wound closure, and IL-8 release. These observations are similar to recently reported effects with a different c-Met TKI in papillary thyroid carcinoma (28). We show for the first time that a clinically relevant c-Met TKI, PF-2341066, can inhibit HNSCC tumor growth in a preclinical animal model.

HGF and/or c-Met overexpression has been described in a wide variety of human tumors including carcinomas of the breast (29), lung (30), head and neck (31, 32), stomach (33), and pancreas (34). Expression of c-Met was found to be even more highly increased in lymph node metastases in HNSCC patients compared with levels in the corresponding primary tumor, which were both elevated compared with normal tissue and unaffected lymph nodes (35). Tumor content of c-Met and HGF has also been associated by others with poor outcome in oral cancer (9, 36, 37). We and others have reported that elevated c-Met and/or HGF expression in human tumors such as lung and breast is often associated with poor clinical outcomes (38–40). Increased HGF serum levels in HNSCC patients were also significantly correlated with higher tumor stage progression (41). The results reported here with regard to clinical outcome represent a very small population compared with previous studies. A trend towards better overall survival was observed only when both HGF and c-Met expression was low in the tumors. Further studies using a larger number of patients will be necessary to establish the clinical usefulness of HGF in either tumor tissue or circulating blood as a biomarker for patients with HNSCC.

In addition to HGF and c-Met overexpression, this pathway can also be activated though genetic alterations such as c-Met–activating mutations. c-Met–activating mutations seem to be rare in all tumor types. Di Renzo et al. identified two somatic constitutively active c-Met mutations in lymph node metastases of HNSCC whereas it was barely detectable in the primary tumors from the same patients, suggesting that these mutations are selected during the metastatic process (42). However, there have been no other reports of c-Met mutations in primary HNSCC tumors.

The HGF pathway has been implicated in clinical resistance to chemotherapy and targeted therapy in HNSCC patients and cell lines. In this regard, Akervall et al. (43) identified low c-Met expression as a predictive factor for favorable response to cisplatin in HNSCC patients. Increased c-Met activation was also recently reported to be correlated with resistance to cetuximab, an EGFR inhibitor, in HNSCC cells (44). Recently, it has been reported that overexpression of cortactin in HNSCC stabilizes c-Met and enhances HGF-induced properties leading to resistance to the EGFR TKI gefitinib (45). Others have reported that EGFR and c-Met are linked via HER3, which has been implicated in clinical resistance to EGFR inhibitors in lung cancer (46). These later findings suggest that c-Met pathway activation can substitute for or act in consort with the EGFR pathway in HNSCC to stimulate growth in the setting of EGFR inhibition.

Molecularly targeted therapies that inhibit the action of pathogenic tyrosine kinases and growth factor pathways have been extensively pursued in recent years. Due to its strong involvement in human cancer, the HGF/c-Met pathway is considered one of the most attractive drug targets for cancer therapy. As such, there are several different therapeutic strategies available to inhibit HGF/c-Met signaling, including small molecule tyrosine kinase inhibitors as well as antagonistic antibodies to either HGF or c-Met that are currently being evaluated in phase I and/or phase II clinical trials for treatment of various types of cancers (47). PF-2341066 is in phase I trials for patients with advanced cancers, including head and neck cancer.⁴

PF-2341066 also has been reported to inhibit the nucleophosmin anaplastic lymphoma kinase (ALK) oncongenic fusion variant of the ALK tyrosine kinase (23). We cannot discount the possibility that the effects that we observed with PF-2341066 are due solely to inhibition of c-Met if these cell populations express ALK. To our knowledge, however, ALK is not expressed in HNSCC cells.

In vitro, we have shown that HGF can induce IL-8 production. Elevated IL-8 tumor levels have been reported to be correlated with more aggressive disease, more recurrences, and shorter disease-free intervals in HNSCC patients (48). In HNSCC tumors, IL-8 and HGF were found to be coexpressed at high levels in a subset of tumors, suggesting that common mechanisms regulate their expression in HNSCC (49). HGF-induced IL-8 production in tumor cells may be an important downstream consequence of the HGF paracrine loop observed in HNSCC tumors.

We report here that close to 80% of primary HNSCC tumors examined in this study express either HGF, c-Met, or both. In addition to HGF binding to c-Met, HGF also binds to heparin binding domains on the cell surface (50). Although the underlying stroma is actually producing HGF, HGF can also be detected in tumor tissue in immunohistochemical staining because it can bind to the tumor cell surface through heparin binding domain interactions. In preparing the tissue microarrays, tissues were cored from tumor epithelium whereas stroma was excluded. Thus, stromal production of HGF could not be evaluated within HNSCC tumors from this tissue microarray. HGF can induce a wide range of biological functions, including invasive growth, in all HNSCC cell lines examined in vitro, and these functions can be inhibited by c-Met TKIs. HGF seems to be necessary for invasion; enhanced invasion due to HGF release by TDFs in coculture can be completely inhibited with a HGF NA. Daly et al. (19) recently reported that invasion of oral squamous cell carcinoma depends on both HGF and stromal derived factor-1 found in conditional medium of stromal fibroblasts. One limitation of the above mentioned study is that fibroblasts from the oral cavity of healthy donors were used whereas our study utilized tumor-derived fibroblasts. In addition, use of a HGF NA in the previous study did not decrease tumor cell invasion in response to fibroblast medium back to control levels whereas we report here complete inhibition with HGF NA. HGF may be the major molecule released by TDFs that promotes tumor invasion. Our data suggest that both an antibody strategy directed at the ligand and a TKI approach directed at the receptor have potential to treat HNSCC. The clinically relevant TKI used here produced significant inhibition of HNSCC xenograft growth and caused a high level of apoptosis and accumulation of cellular debris within tumor xenografts.

Head and neck cancer has been particularly resistant to systemic therapies, and new treatments such as the EGFR monoclonal antibody cetuximab have shown little improvement. Due to either intrinsic or acquired resistance to EGFR inhibitors, the clinical utility of this class of drugs is limited. Thus, new strategies are greatly needed to combat this disease. Together these results suggest that therapeutic targeting of the HGF/c-Met pathway may offer new possibilities in the treatment of HNSCC patients either alone or in combination with other treatment strategies. Future research should focus on the identification of a population of patients most likely to benefit from this treatment. The rational design of combinations of molecularly targeted agents which may have a survival benefit to single-agent therapies in the treatment of head and neck cancers is necessary.

No potential conflicts of interest were disclosed.